{"title":"Una comparación empírica de algoritmos de aprendizaje automático versus aprendizaje profundo para la detección de noticias falsas en redes sociales","authors":"Luis Rojas Rubio, Claudio Meneses Villegas","doi":"10.4067/s0718-33052022000200403","DOIUrl":null,"url":null,"abstract":"Las redes sociales se han convertido en uno de los principales canales de información del ser humano debido a la inmediatez e interactividad social que ofrecen, permitiendo en algunos casos publicar lo que cada usuario considere pertinente. Esto ha traído consigo la generación de noticias falsas o Fake News, publicaciones que solo buscan generar incertidumbre, desinformación o sesgar la opinión de los lectores. Se ha evidenciado que el ser humano no es capaz de identificar en su totalidad si un artículo es realmente un hecho o bien una Fake News, debido a esto es que surgen modelos que buscan caracterizar e identificar artículos basados en minería de datos y machine learning. Este artículo compara empíricamente distintos esquemas de machine learning y deep learning en la tarea de identificar fake news. Para ello se utilizan conjuntos de datos extraídos desde el estado del arte. Los resultados obtenidos en base a la técnica de muestreo utilizado y la representación vectorial Tf-Idf del corpus, indica una mejora significativa en el accuracy en contraste a los resultados obtenidos en el estado del arte considerando el repositorio FakeNewsNet.Alternate :Social networks have become one of the leading information channels for human beings due to the immediacy and social interactivity they offer, allowing, in some cases, to publish what each user considers relevant. This usage has brought with it the generation of false news or Fake News, publications that only seek to generate uncertainty, misinformation, or skew the readers' opinion. It has been shown that the human being is not able to fully identify whether an article is actually a fact or a Fake News;due to this, models that seek to characterize and identify articles based on data mining and machine learning emerge. This article empirically compares different machine learning and deep learning schemes to identify fake news;data sets extracted from state of the art are used to accomplish this. The results obtained based on the sampling technique used and the Tf-Idf vector representation of the corpus indicate a significant improvement in accuracy in contrast to the results obtained in the state of the art considering the FakeNewsNet repository.","PeriodicalId":40015,"journal":{"name":"Ingeniare","volume":"1 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2022-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ingeniare","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.4067/s0718-33052022000200403","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"Engineering","Score":null,"Total":0}
引用次数: 0
Abstract
Las redes sociales se han convertido en uno de los principales canales de información del ser humano debido a la inmediatez e interactividad social que ofrecen, permitiendo en algunos casos publicar lo que cada usuario considere pertinente. Esto ha traído consigo la generación de noticias falsas o Fake News, publicaciones que solo buscan generar incertidumbre, desinformación o sesgar la opinión de los lectores. Se ha evidenciado que el ser humano no es capaz de identificar en su totalidad si un artículo es realmente un hecho o bien una Fake News, debido a esto es que surgen modelos que buscan caracterizar e identificar artículos basados en minería de datos y machine learning. Este artículo compara empíricamente distintos esquemas de machine learning y deep learning en la tarea de identificar fake news. Para ello se utilizan conjuntos de datos extraídos desde el estado del arte. Los resultados obtenidos en base a la técnica de muestreo utilizado y la representación vectorial Tf-Idf del corpus, indica una mejora significativa en el accuracy en contraste a los resultados obtenidos en el estado del arte considerando el repositorio FakeNewsNet.Alternate :Social networks have become one of the leading information channels for human beings due to the immediacy and social interactivity they offer, allowing, in some cases, to publish what each user considers relevant. This usage has brought with it the generation of false news or Fake News, publications that only seek to generate uncertainty, misinformation, or skew the readers' opinion. It has been shown that the human being is not able to fully identify whether an article is actually a fact or a Fake News;due to this, models that seek to characterize and identify articles based on data mining and machine learning emerge. This article empirically compares different machine learning and deep learning schemes to identify fake news;data sets extracted from state of the art are used to accomplish this. The results obtained based on the sampling technique used and the Tf-Idf vector representation of the corpus indicate a significant improvement in accuracy in contrast to the results obtained in the state of the art considering the FakeNewsNet repository.
由于社交网络提供的即时性和互动性,社交网络已经成为人类信息的主要渠道之一,在某些情况下允许发布每个用户认为相关的内容。这导致了假新闻的产生,这些出版物只是试图制造不确定性、错误信息或扭曲读者的观点。事实证明,人类无法完全识别一篇文章是真实的事实还是假新闻,正因为如此,基于数据挖掘和机器学习的模型试图描述和识别文章。本文实证比较了不同的机器学习和深度学习方案在识别假新闻任务中的应用。为了实现这一目标,我们使用了最先进的数据集。基于所使用的抽样技术和Tf-Idf向量表示的结果表明,与考虑FakeNewsNet存储库的最新结果相比,准确性有了显著提高。另一种选择:社交网络已经成为人类的主要信息渠道之一,因为它们提供的即时性和社交互动性,在某些情况下,允许发布每个用户认为相关的内容。你的with it the generation of This usage false news or举news,出版物that only失所to generate对立,misinformation, or skew the读者意见。人们发现,人类无法完全识别一篇文章是事实还是假新闻;因此,出现了试图描述和识别基于数据挖掘和机器学习的文章的模型。本文从经验上比较了不同的机器学习和深度学习方案来识别假新闻;从最先进的技术中提取的数据集用于实现这一点。根据所使用的抽样技术和Tf-Idf语料库向量表示所取得的结果表明,与考虑到假新闻网储存库的最新情况所取得的结果相比,其准确性有了显著提高。
期刊介绍:
Ingeniare. Revista chilena de ingeniería is published periodically, is printed in three issues per volume annually, publishing original articles by professional and academic authors belonging to public or private organisations, from Chile and the rest of the world, with the purpose of disseminating their experiences in engineering science and technology in the areas of Electronics, Electricity, Computing and Information Sciences, Mechanical, Acoustic, Industrial and Engineering Teaching. The abbreviated title of the journal is Ingeniare. Rev. chil. ing. , which should be used in bibliographies, footnotes and bibliographical references and strips.